In recent years academic research has focused on understanding and modeling the survey response process. This paper examines an understudied systematic response tendency in surveys: the extent to which observed responses are subject to state dependence, i.e., response carryover from one item to another independent of specific item content. We develop a statistical model that simultaneously accounts for state dependence, item content, and scale usage heterogeneity. The paper explores how state dependence varies by response category, item characteristics, item sequence, respondent characteristics, and whether it becomes stronger as the survey progresses. Two empirical applications provide evidence of substantial and significant state dependence. We find that the degree of state dependence depends on item characteristics and item sequence, and it varies across individuals and countries. The article demonstrates that ignoring state dependence may affect reliability and predictive validity, and it provides recommendations for survey researchers.

Additional Metadata
Keywords item response theory, response process, scale usage, state dependence, survey response models, surveys
Persistent URL dx.doi.org/10.1287/mksc.1120.0722, hdl.handle.net/1765/37425
Series ERIM Top-Core Articles
Journal Marketing Science: the marketing journal of INFORMS
Citation
de Jong, M.G, Lehmann, D.R, & Netzer, O. (2012). State-Dependence Effects in Surveys . Marketing Science: the marketing journal of INFORMS, 31(5), 838–854. doi:10.1287/mksc.1120.0722